I kind of agree that
yaml isn’t the best way to have all those graphs. But we haven’t found yet a better way to do it. What we wanted is to version control them, and treat them as any other code (PRs, etc).
Our way of working is having two redash instances, in one we work and experiment, then we export the queries and graphs in
yaml we use those as templates to be fed through a templating lib to the production redash. In that production there is only one user which can write queries (the one used by redpush from CI) and the other users have only read access. Hence the only way to change it is through the PR process.
And regarding the question, I think (I haven’t tested it though) if you create queries without graphs, they will work fine. Redash automatically adds a visualization that is the table view. And if you update one, if no visualizations are in the
yaml redpush should not change the ones already available there. But as said I haven’t tested it.